This dataset contains GPS-based locomotion and spatial behaviour data of horses monitored under different housing systems at two Agroscope research sites (in Switzerland) between December 2022 and July 2024 (12 different assessment periods). GPS tracking was conducted using collar-mounted Qstarz loggers with a 10-second recording interval, providing continuous movement data for individual horses across multiple observation periods. The dataset is organised into 12 folders representing individual GPS assessment periods. Each folder contains raw GPS data (GPX files) for individual horses. Spatial utilisation layers and associated outputs (e.g. functional area classifications) are not included in the repository, as they are not central to the study objectives and can be fully reproduced using the provided scripts. Processed datasets include GPS data prepared for movement analysis (derived metrics such as daily distance travelled per horse). These are provided in a separate folder ("Summary_tables_per_assessment_period") and allow reuse without the need to reprocess raw GPS data. The repository includes: raw and combined GPS datasets (RDS format), spatially processed datasets with polygon classification (RDS, SHP), intermediate geospatial layers (GPKG, raster), aggregated movement metrics (e.g. daily distance travelled), final statistical datasets used for modelling, fully annotated R scripts enabling complete reproduction of all analyses. Methods overview: GPS data were processed in R using a fully scripted workflow. Coordinates were transformed into the Swiss reference system (EPSG:2056) and filtered to retain only positions within the study area. Movement metrics were calculated based on stepwise Euclidean distances between consecutive GPS fixes, applying a maximum interval threshold of 30 seconds. Spatial utilisation was quantified by assigning GPS points to predefined functional zones derived from QGIS polygons. Time spent per functional area and movement patterns were subsequently aggregated at the daily and individual level. Statistical analyses were performed using linear mixed-effects models to evaluate the influence of shoeing status and environmental variables (e.g. Temperature–Humidity Index) on daily locomotion. Reproducibility: All data and code necessary to reproduce the analyses are publicly available in this repository. All analyses were conducted in R using a fully reproducible workflow. Movement metrics were calculated based on Euclidean distances between consecutive GPS fixes (≤30 s intervals), and spatial utilisation was derived from polygon-based classification of GPS positions. Raw GPX files are provided to allow full reproducibility of data processing and analysis. Notes: GPS accuracy is approximately 1–3 m under open-sky conditions. Reduced accuracy may occur near sheltered structures and was not corrected analytically.